JP2006180269A - Image processing apparatus, image processing method, imaging apparatus, program, and recording medium - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an image processing apparatus, an image processing method, an imaging apparatus, a program, and a recording medium for reducing the noise of an imaged image under natural light without using a lighting fixture for affecting the image. <P>SOLUTION: An imaging section 2 images a visible light image Visible and an infrared ray image Infr. In the case of imaging the images, the imaging section 2 images the images under natural light without using a particular lighting fixture. A low pass filter 33 eliminates noise in the visible light image Visible and a high pass filter 34 extracts an edge of the infrared ray image Infr. An output image Base from the low pass filter 34 has an advantage of correct coloring of the image and a disadvantage of an unclear detailed part of the image. An output image Edge from the high pass filter 34 keeps the edge of the infrared ray image Infr and its detailed parts. An image composite section 35 composes the images Base and Edge to obtain an output image OUT having both advantages of both the images. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

  The present invention relates to an image processing apparatus, an image processing method, an imaging apparatus, a program, and a recording medium that remove noise while preserving image edges.

  When an image is captured by a camera, the color of the image greatly depends on lighting. For example, if an image of a subject illuminated by candlelight is captured, an orange-colored image is obtained, and if an image of a subject illuminated by moonlight is imaged, a bluish-colored image is obtained. Even in the same place, a completely different image is captured depending on the lighting at the time of imaging.

  When capturing images, it is possible to reproduce the correct color by using natural light (hereinafter referred to as natural light) without using a special lighting device such as a flash. When the image is taken, there is a problem that the exposure is small and the image contains noise. On the other hand, when using a flash, the amount of exposure increases and the edges and details are captured clearly, but the color tone is not accurate, and shadows and highlights that do not actually exist may occur. appear.

  Conventionally, in order to solve such a problem, an image processing apparatus 11 as shown in FIG. 12 has been proposed. The image processing apparatus 11 includes two low-pass filters 12 and 13, one high-pass filter 14, two image synthesis units 15 and 17, and one shadow extraction unit 16.

  The low-pass filter 13 is a cross bilateral filter, and detects noise from an image using a flash (hereinafter referred to as a flash image) and removes noise from an image not using the flash (hereinafter referred to as a natural light image). The high pass filter 14 extracts the edge of the flash image. Edge extraction is performed by dividing each pixel value of the natural light image by the flash image.

  The image synthesis unit 15 synthesizes the natural light image from which noise has been removed by the low-pass filter 13 and the flash image from which the edge has been extracted by the high-pass filter 14 to generate a synthesized image Comp. For image synthesis, a process of multiplying the pixel value of the low-pass filter 13 by the pixel value of the output image of the high-pass filter 14 is performed. The composite image Comp has the advantages of a flash image and a natural light image, is accurate in color, and has less noise.

  The low-pass filter 12 removes noise from the natural light image. A bilateral filter is used as the low-pass filter 12. The bilateral filter is a filter that performs edge detection and noise removal of an image using a single image.

  The shadow extraction unit 16 extracts the difference between the two images of the flash image and the natural light image, and evaluates the probability that the image has changed due to highlight or shadow. The image synthesis unit 17 performs weighted addition of the output image from the low-pass filter 12 and the output image Comp of the image synthesis unit 15 based on the evaluation result of the shadow extraction unit 16. The image synthesizing unit 17 increases the weight of the image in a portion where there is a high possibility that a shadow or highlight has occurred due to the presence or absence of a flash, and increases the weight of the image in a portion where the possibility is low. The image synthesizing unit 17 generates an output image Out by removing unnecessary shadows and highlights from the synthesized image Comp.

  As described above, the conventional image processing apparatus 11 generates two images, that is, an image obtained by combining the edges of the flash image and an image from which noise has been removed without referring to the edges of the flash image. The pixel of the output image of the high-pass filter 12 is increased for pixels that are highly likely to have highlights, and the coefficient of the output image Comp of the image composition unit 17 is increased for pixels that are less likely to cause shadows or highlights due to flash. By doing so, an image in which the edge of the flash image and the color of the natural light image are optimally blended can be obtained (for example, Non-Patent Document 1).

Georg Petschnig Etal, Digital Photography with Flash and No-Flash Image Pairs, acm Transaction on Graphics, Vol. 23, Number 3, pp. 664-672, August 2004

  As described above, when a flash is used, the edges and details of an image become clear, but shadows and highlights that do not exist in natural light may occur. The image processing apparatus 11 wants to leave only edges and details while removing shadows and highlights. However, it is not easy to distinguish between these, and it requires a large calculation cost.

  The image processing apparatus 11 requires two images, a flash image and a natural light image. Since a flash image and a natural light image cannot be captured at the same time, there is a problem that it cannot be applied to a moving image or a moving subject. There is also a problem that a flash image cannot be obtained at a place where the use of the flash is prohibited.

  The present invention has been made in view of the above-described problems, and an image processing apparatus and image processing that reduce noise in an image captured under natural light without using a lighting fixture that affects the image. It is an object to provide a method, an imaging apparatus, a program, and a recording medium.

  In order to achieve the above-described object, an image processing device to which the present invention is applied includes an image acquisition unit that acquires a visible light image, an invisible light image obtained by imaging the same subject as the visible light image, and an invisible light image. And noise reduction means for reducing noise in the visible light image.

  An imaging apparatus to which the present invention is applied includes a visible light image capturing unit that captures a visible light image based on a first spectral characteristic that is mainly sensitive to a visible light image, and a sensitivity that is mainly sensitive to an invisible light image. An invisible light image pick-up means for picking up an invisible light image based on a certain second spectral characteristic, an aberration correction means for correcting an aberration between the visible light image and the invisible light image, and a visible light image using the invisible light image Noise reduction means for reducing the noise.

  An image processing method to which the present invention is applied includes an image acquisition step of acquiring a visible light image and an invisible light image obtained by imaging the same subject as the visible light image, and reducing noise in the visible light image using the invisible light image. Noise reduction step.

  The program to which the present invention is applied is a program that causes a computer to execute a predetermined process, and obtains a visible light image and an invisible light image that corresponds to the visible light image and is captured with the same number of pixels as the visible light image. An acquisition step and a noise reduction step of reducing noise in the visible light image using the invisible light image.

  A recording medium to which the present invention is applied is a visible medium that records a program that causes a computer to execute a predetermined process, and is invisible corresponding to the visible light image and the same number of pixels as the visible light image. A program having an image acquisition step of acquiring a light image and a noise reduction step of reducing noise of a visible light image using an invisible light image is recorded.

  According to the present invention, in order to reduce the noise of an invisible light image using a visible light image, the noise of an image captured under natural light can be reduced without using a lighting fixture that affects the image. Can do.

  Hereinafter, an imaging apparatus to which the present invention is applied will be described with reference to the drawings. FIG. 1 shows the configuration of the imaging apparatus 1. The imaging device 1 includes an imaging unit 2 that captures both a visible light image Visible and an infrared light image Infr, an image processing unit 3 that removes noise from the visible light image Visible, and a memory that is a storage area for images and data. 4, a system control unit 8 that performs image output to an LCD (Liquid Crystal Display) 5, data transmission / reception with an external recording device 10 via an interface such as a serial interface 6 or a USB (Universal Serial Bus) 7, and an imaging device And a signal processing unit 9 that performs AGC (Automatic Gain Control) and CDS (Correlated Double Sampling) on the image input from 21 and outputs the image to the image processing unit 3.

  The imaging unit 2 outputs RGB images of a visible light image Visible and an infrared light image Infr. The imaging unit 2 controls, for example, an imaging element 21 that forms a CCD (Charge Coupled Device) or a CMOS (Complementary Metal-Oxide Semiconductor), a motor 23 that performs focusing of the lens 22 and switching of a shutter, and a motor 23. And a driver 24.

  The image pickup device 21 includes an image pickup device 21 for picking up an infrared image Infr and an image pickup device 21 for picking up a visible light image Visible. The image sensor 21 that captures the infrared light image Infr and the image sensor 21 that captures the visible light image Visible have the same number of pixels and the same angle of view, and are exposed for the same time at the same time. Note that the imaging element 21 may be separate for visible light and infrared light, or the output of one imaging element 21 may be dispersed.

  The image processing unit 3 generates an output image Out from which noise is removed while preserving the edge of the visible light image Visible. FIG. 2 shows the configuration of the image processing unit 3. The image processing unit 3 includes a gain adjusting unit 31 that adjusts the gain of the visible light image Visible, a low-pass filter 33 that removes noise from the visible light image Visible, and a high-pass filter that extracts edges and details of the infrared light image Infr. 34 and an image synthesis unit 35 that synthesizes a basic image Base that is an output image of the low-pass filter 33 and a detailed image Edge that is an output image of the high-pass filter 34.

  The gain adjustment unit 31 adjusts the gain to increase the pixel value of the visible light image Visible imaged darkly due to insufficient exposure so that the pixel value is close to that of the image imaged with appropriate exposure. As a gain adjustment method, there are a method of multiplying a pixel value of a visible light image by a constant, a gamma correction based on an exponential function, an arbitrary gain adjustment method based on a polynomial function, and the like. The maximum pixel value after adjustment is limited.

  FIG. 3 shows the configuration of the low-pass filter 33. The low-pass filter 33 includes an X edge detection unit 41 that detects an edge in the X direction (width direction) of the infrared light image Infr, and a Y edge detection that detects an edge in the Y direction (height direction) of the infrared light image Infr. Unit 42, an X evaluation value storage unit 43 that stores edge evaluation values in the X direction, a Y evaluation value storage unit 44 that stores edge evaluation values in the Y direction, and three X low-pass filters that remove edges in the X direction 45a, 45b, 45c, three Y low-pass filters 46a, 46b, 46c for removing edges in the Y direction, three X comparison units 47a, 47b, 47c for comparing X edge evaluation values and threshold values, and Y edges Three Y comparison units 48a, 48b, and 48c that compare the evaluation value and the threshold value are provided.

  The X low pass filter is a 5 × 1 tap FIR low pass filter in the X direction. The noise in the X direction of the visible light image is removed by the X low-pass filter. The Y low-pass filter is a 1 × 5 tap FIR low-pass filter in the Y direction. The Y low-pass filter removes noise in the Y direction of the visible light image.

  The X edge detection unit 41 is a 4 × 4 FIR filter that detects an edge in the X direction, and the Y edge detection unit 42 is a 4 × 4 FIR filter that detects an edge in the Y direction. The X evaluation value storage unit 43 applies an absolute value calculation to the filter result of the X edge detection unit 42 to calculate an edge evaluation value, and stores this value as an X edge image. The Y evaluation value storage unit 44 applies an absolute value calculation to the filter result of the Y edge detection unit 42 to calculate an edge evaluation value, and stores this value as a Y edge image.

  The X comparison unit 47 and the Y comparison unit 48 compare the edge evaluation value with a predetermined threshold value. The threshold value n is ½ of the maximum value of the edge evaluation value. The X comparison unit 47a compares the edge evaluation value in the X direction with the threshold value n, the X comparison unit 47b compares the edge evaluation value in the X direction with the threshold value n / 2, and the X comparison unit 47c performs the edge evaluation in the X direction. The value is compared with the threshold value n / 4. The Y comparison unit 48a compares the edge evaluation value in the Y direction with the threshold value n, the Y comparison unit 48b compares the edge evaluation value in the Y direction with the threshold value n / 2, and the Y comparison unit 48c The edge evaluation value is compared with the threshold value n / 4.

  The low-pass filter 33 is composed of three stages: a level 1 filter 49a, a level 2 filter 49b, and a level 3 filter 49c. The level of the low-pass filter 33 depends on the threshold value in the X comparison unit 47 and the Y comparison unit 48. The low-pass filter with threshold n is called level 1 filter 49a, the low-pass filter with threshold n / 2 is called level 2 filter 49b, and the low-pass filter 49c with threshold n / 4 is called level 3 filter.

  The visible light image Visible is first output to the X comparison unit 47a. The X comparison unit 47 a compares the edge evaluation value in the X direction stored in the X evaluation value storage unit 43 with the threshold value n. When the edge evaluation value is smaller than the threshold value n, the X comparison unit 47a outputs the visible light image Visible to the X low-pass filter 45a, and when the edge evaluation value is greater than or equal to the threshold value n, the X comparison unit 47a compares the visible light image Visible with the Y comparison. To the unit 48a. The comparison unit 48b compares the edge evaluation value in the Y direction stored in the Y evaluation value storage unit 44 with the threshold value n. When the edge evaluation value is smaller than the threshold value n, the comparison unit 48b outputs the visible light image Visible to the Y low-pass filter 46a, and when the edge evaluation value is equal to or larger than the threshold value n, the visible light image Visible is output to the next X. It outputs to the comparison part 47b.

  Similarly, the X comparison unit 47b, the Y comparison unit 46b, the X comparison unit 47c, and the Y comparison unit 48c compare the edge evaluation value with the threshold value. If the threshold value is smaller than the threshold value, the visible light image Visible is displayed on the subsequent low-pass filter. If it is greater than or equal to the threshold value, the visible light image Visible is output to the next comparison unit.

  FIG. 4 shows the operation of the low-pass filter 33. The low-pass filter 33 first inputs the infrared light image Visible (step S1). The X edge detector 41 detects an edge in the X direction existing in the infrared light image Infr. The X evaluation value storage unit 43 calculates an edge evaluation value in the X direction by applying a predetermined absolute value calculation to the edge in the X direction, and stores the calculated edge evaluation value as an X edge image (step S2). The Y edge detector 42 detects an edge in the Y direction existing in the infrared light image Infr. The Y evaluation storage unit 44 calculates an edge evaluation value in the Y direction by applying a predetermined absolute value calculation to the edge in the Y direction, and stores the calculated edge evaluation value as a Y edge image (step S3).

  When the visible light image Visible is input from the gain adjustment unit 31 (step S4), the low-pass filter 33 applies the input visible light image Visible to the level 1 filter 49a and performs processing in the X direction (step S5).

  FIG. 5 shows processing in the X direction of the level 1 filter 49a. In the X-direction processing of the level 1 filter 49a, first, the Y-direction coordinate Y is initialized to 0 (step S11), and the X-direction coordinate X is initialized to 0 (step S12). The X comparison unit 47 a inputs an X evaluation value image from the X evaluation value storage unit 43. The X comparison unit 47a compares the edge evaluation value in the X direction at the coordinates (X, Y) of the X evaluation value image with the threshold value n. When the edge evaluation value at the coordinates (X, Y) is smaller than the threshold value n (step S13; YES), the X low-pass filter 45a is applied to the visible light image at the coordinates (X, Y) (step S14). On the other hand, when the edge evaluation value of the coordinates (X, Y) is smaller than the threshold value n (step S13; NO), the process proceeds to step S15. In step S15, the X coordinate is incremented by 1 (step S15). The value of the X coordinate is compared with the width of the visible light image Visible. If the coordinate X is smaller than the width of the visible light image Visible (step S16; YES), the process proceeds to step S13. On the other hand, if the coordinate X is larger than the width of the visible light image Visible (step S16; NO), the coordinate Y is incremented by 1 (step S17). In step S18, the value of the Y coordinate is compared with the height of the visible light image Visible. If the Y coordinate is smaller than the height of the visible light image Visible (step S18; YES), the process proceeds to step S12. On the other hand, if the Y coordinate is larger than the height of the visible light image Visible (step S18; NO), the processing of the level 1 filter in the X direction is terminated. The level 1 filter compares the edge evaluation value of each pixel constituting the visible light image with a threshold value n. If the edge evaluation value is smaller than the threshold value n, the X low-pass filter 45a is applied to remove noise in the X direction.

  In the level 1 filter, when the processing by the level 1 filter 49a in the X direction is completed, the processing by the level 1 filter 49a in the Y direction is performed. Here, substantially the same processing as the level 1 filter 49a in the X direction shown in FIG. 5 is performed. In the level 1 filter 49a in the Y direction, the Y edge image is used as the edge evaluation image instead of the X edge image, and the Y low pass filter 46a is used as the low pass filter instead of the X low pass filter 45a (step S6).

  The level 1 filter 49a outputs the visible light image Visible to the level 2 filter 49b when its processing is completed. The level 2 filter 49b first performs processing in the X direction (step S7). When processing in the X direction is completed, processing in the Y direction is performed (step S8). When the output image of the level 2 filter 49b is input, the level 3 filter 49c performs processing in the X direction (step S9), and when the processing in the X direction is completed, performs processing in the Y direction (step S10). The level 2 filter and the level 3 filter perform the same processing except that the threshold values are different.

  In the low-pass filter 33, the number of times of filtering increases as the pixel with a lower edge evaluation value, and the number of times of filtering decreases as the pixel with a higher edge evaluation value. That is, since a pixel with a high edge evaluation value has a small number of times of filtering, the edge is stored, and a pixel with a low edge evaluation value has a high number of times of filtering, so that noise is removed. A filter having such a function is called an edge preserving filter. The types of edge preserving filters include bilateral filters and cross bilateral filters in addition to those shown in FIG. These filters may be used as the low-pass filter 33.

  The low-pass filter 33 outputs an image from which noise of the visible light image Visible is removed. This image is called a basic image Base. The basic image Base has the advantage that the color of the image is correct and the disadvantage that the edges and details are not clear and give a blurred impression.

  The high pass filter 34 extracts an edge portion of the infrared light image Infr. FIG. 6 shows an example of the high pass filter 34. The high pass filter 34 in FIG. 6 is a two-dimensional FIR filter. The high pass filter 34 includes a low pass filter 71 and a division unit 72. The low-pass filter 71 is, for example, an edge-preserving low-pass filter. The low-pass filter 71 removes noise from the visible light image Visible and outputs this image to the division unit 72. The division unit 72 divides the output of the low-pass filter 71 from the infrared light image Infr, and extracts a high-pass component of the infrared light image Infr. In the image output from the high-pass filter 34, edges and details of the infrared light image Infr are stored. This image is called a detailed image Edge.

  The image composition unit 35 generates a composite image obtained by multiplying the basic image Base and the detailed image Edge. This image is the output image OUT of the image processing unit 3. The output image OUT is an image obtained by combining the advantages of the two images of the basic image Base and the detailed image Edge, and has a characteristic that the color portion is correct and the detailed portion is clear.

  As described above, in the imaging apparatus 1 to which the present invention is applied, the basic image Base from which the noise of the visible light image Visible is removed and the detailed image Edge from which the edge and the detailed portion of the infrared light image Infr are extracted are synthesized. Accordingly, it is possible to obtain an output image OUT including an edge portion and a detailed portion attenuated by noise removal while removing noise of the visible light image Visible.

  In addition, since the infrared light image Infr can be captured at the same time as the visible light image Visible, the imaging time does not change, and a moving image or a moving subject can be processed.

  In the conventional image processing apparatus, since the edge is extracted from the flash image, the calculation cost for removing shadows and highlights caused by the difference in illumination conditions has increased, but the infrared light image Infr is visible light. Since the image can be captured under the same illumination condition as the image Visible, it is not necessary to correct the difference in the illumination condition.

  Further, the conventional image processing apparatus has a problem that pixels in a detailed portion are discarded when a detailed portion or an edge is erroneously determined as a shadow or a highlight. Since the infrared light image Infr can be captured under the same illumination conditions as the visible light image Visible, no shadows or highlights are generated due to differences in the illumination conditions, and there is no possibility that necessary pixels are discarded.

  Next, a first modification of the image processing unit 3 will be described with reference to FIG. The image processing unit 50 performs edge detection using the luminance of the visible light image Visible. Since the luminance has one variable, the calculation cost is lower than when performing edge detection from three RGB variables. Since humans are generally sensitive to luminance and not sensitive to color components, even edge detection from luminance is sufficiently effective.

  FIG. 7 shows the configuration of the image processing unit 50. The image processing unit 50 includes a gain adjustment unit 52, a matrix unit 53, a color low-pass filter 54, a luminance low-pass filter 55, a high-pass filter 56, an image composition unit 57, and an inverse matrix unit 58.

  The gain adjustment unit 52 adjusts the gain to increase the pixel value of a visible light image captured darkly due to insufficient exposure so that the pixel value is close to that of an image captured with appropriate exposure. As a gain adjustment method, there are a method of multiplying a pixel value of a visible light image by a constant, a gamma correction based on an exponential function, an arbitrary gain adjustment method based on a polynomial function, and the like. The maximum pixel value after adjustment is limited.

  The matrix unit 53 performs matrix conversion on the RGB image to convert it into color images Cb and Cr and a luminance image Yd. The color low-pass filter 54 removes noise from the color images Cb and Cr. As the color low-pass filter 54, for example, a bilateral filter is used. The bilateral filter is a filter that performs edge detection and noise removal from one image.

  The luminance low-pass filter 55 removes noise from the luminance image Yd. As the luminance low-pass filter 55, for example, a cross bilateral filter is used. The cross bilateral filter is a filter that removes noise from the filter target image while preserving edges detected from the image for edge detection. Here, an edge is detected from the infrared light image Infr to remove noise from the luminance image Yd. An image output from the luminance low-pass filter 55 is referred to as a basic image Base. The basic image Base has the advantage that the luminance of the image is correct and the disadvantage that the edges and details are not clear and give a blurred impression.

  The high pass filter 56 extracts an edge portion of the infrared light image Infr. In the image output from the high-pass filter 56, the edges and details of the infrared light image Infr are stored. This image is called a detailed image Edge.

  The image composition unit 57 generates a composite image obtained by multiplying the basic image Base and the detailed image Edge. This composite image is an image in which the advantages of the basic image Base and the detailed image Edge are combined, and the luminance is correct and the detailed portions and edges are clear. The inverse matrix unit 58 performs inverse matrix conversion on the composite image and converts the luminance image into an RGB image. This image is the output image OUT of the image processing unit.

  The image processing unit 50 separates the RGB image into color images Cb and Cr and a luminance image Yd, and performs edge detection only on the luminance image Yd. Humans are generally sensitive to luminance and not sensitive to color components. The calculation cost can be reduced by filtering only the luminance instead of the three variables of RGB.

  Next, a second modification of the image processing unit 3 will be described. As shown in FIG. 8, the image processing unit 30 includes an aberration correction unit 32 that corrects the aberration of the infrared light image. The configuration other than the aberration correction unit 32 is the same as that of the image processing unit 3. The same components as those in the image processing unit 3 are denoted by the same reference numerals. Description of these components is omitted.

  The aberration correction unit 32 corrects an aberration caused by a difference in wavelength between infrared light and visible light. FIG. 9 shows the configuration of the aberration correction unit 32. The aberration correction unit 32 includes a matrix unit 61, an error calculation bilinear scaler 62, an aberration correction bilinear scaler 63, a parameter calculation unit 64, and an error calculation unit 65.

  The matrix unit 61 generates a luminance image Yd of the input visible light image Visible. The error calculating bilinear scaler 62 generates five scale-converted images and distortion-converted images based on the scale value and distortion value output from the parameter calculation unit 64.

  The error calculation unit 65 compares the scale-converted image, the infrared light image, the distortion-converted image, and the infrared light image, and calculates an error value of these images. This error value is a PSNR value (noise mixing amount; Peak Signal To Noise Ratio). The parameter calculation unit 64 refers to the PSNR value calculated by the error calculation unit 65 and optimizes the scale value and the distortion value. The aberration correction bilinear scaler 63 corrects the aberration of the visible light image Visible using the scale value and the distortion value optimized by the parameter calculation unit 64.

  FIG. 10 shows the operation of the aberration correction unit 32. The aberration correction unit 32 receives the visible light image Visible and the infrared light image Infr. This visible light image Infr is an RGB image (step S21). The matrix unit 61 generates the luminance image Yd by multiplying the visible light image Visible by the matrix (step S22). The parameter calculation unit 64 initializes the maximum value, current value, and minimum value of the scale value, and the maximum value, current value, and minimum value of the distortion value (step S23).

  The parameter calculation unit 64 receives the PSNR value from the error calculation unit 65, and obtains a scale value that maximizes the PSNR value (step S24).

FIG. 11 shows the procedure for calculating the scale value. The parameter calculation unit 64 has five parameters of a maximum value S ₁ of the scale value, an intermediate value S ₂ between the maximum value and the current value, a current value S ₃ , an intermediate value S ₄ between the maximum value and the current value, and a minimum value S _5. prepare. The error detection bilinear scaler 62 performs scale conversion of the luminance image Yd using each of these five scale values. Thus, the five luminance image _Yd 1 _{~Yd 5} which correspond to the five parameters is generated (step S31).

The error calculation unit 65 compares the luminance images Yd _{1 to} Yd ₅ and the infrared light image Infr to obtain the PSNR value. When each luminance image and the infrared light images Yd _{1 to} Yd ₅ are compared, five PSNR values are calculated (step S32). When the PSNR value of the luminance image Yd ₁ scale-converted with the maximum value S ₁ and the infrared light image Infr is the maximum (step S 33; YES), the parameter calculation unit 64 sets the current value S ₃ to the maximum value S ₁ . replacement, replacing the maximum value _{S 1} with the value obtained by subtracting the minimum value _{S 5} from twice the value of the maximum value _{S 1} (step S34).

If PSNR value in the intermediate value of the maximum values _{S 1} and the current value _{S 3} is maximum (step S35; YES), the parameter computing section 64 replaces the minimum value _{S 5} as the current value _{S 3,} as the current value _{S 3} maximum values _{S 1} and replace with an intermediate value of the current value _{S 3} (step S36).

If the PSNR value is largest for the current value _{S 3} (step S37; YES), the parameter computing section 64 replaces the maximum value _{S 1} at the current value _{S 3,} the current value _{S 3} the current value _{S 3} and the minimum value S The intermediate value of ₅ is replaced (step S38).

If PSNR value in the middle of the minimum value _{S 5} as the current value _{S 3} is maximum (step S39; YES), the parameter computing section 64 replaces the current value _{S 3} at the minimum value _{S 5,} the minimum the minimum value _{S 5} It replaces 2 times the value _{S 5} with the value obtained by subtracting the maximum value _{S 1} (step S40).

If the PSNR value is largest for the minimum value _{S 5} (step S33, step S35, step S37, step S39; NO), the parameter computing section 64 replaces the current value _{S 3} at the minimum value _{S 5,} the minimum value _{S 5} replacing the maximum value _{S 5} from twice the minimum value _{S 5} (step S41). In this way, the optimization of the scale value is completed.

  Next, the parameter calculation unit 64 optimizes the distortion value. Distortion value optimization is the same processing as scale value optimization processing. In the optimization of the distortion value, the parameter changes from the scale value to the distortion value (step S25). When the optimization of the scale value and the distortion value is completed, the parameter calculation unit 64 compares the improvement amount of the PSNR value with a predetermined threshold value. When the improvement amount of the PSNR value is lower than the predetermined threshold (step S26; NO), the process proceeds to step S25 and the scale value and the distortion value are optimized again.

  On the other hand, when the improvement amount of the PSNR value is higher than the predetermined threshold (step S26; YES), the parameter calculation unit 64 outputs the current scale value and distortion value to the aberration correction bilinear scaler 63. The aberration correcting bilinear scaler 63 performs aberration correction of the visible light image Visible using the current scale value and distortion value (step S27), and outputs the corrected image to the high-pass filter 56 (step S28).

  By providing the aberration correction unit 32, it is possible to correct a difference in aberration caused by a difference in wavelength between visible light and infrared light. Thereby, the image imaged by visible light image Visible and the image imaged by infrared light image Infr correspond.

  In general, the refractive index of infrared light is high and slightly enlarged. When the visible light image Visible is corrected, a large image is obtained, but distortion is also increased. Since the size of the image and the size of the distortion are in a trade-off relationship, the image to be corrected differs depending on which is prioritized.

  The filtering process and the aberration correction process in the image processing unit 3 may be executed based on a control program. Such a control program is recorded in the firmware of the imaging apparatus 1. The control program may be acquired via a recording medium recorded in a format readable by the external recording device 9. As a recording medium for recording the control program, a magnetic reading type recording medium (for example, magnetic tape, flexible disk, magnetic card), an optical reading type recording medium (for example, CD-ROM, MO, CD-R, DVD) A semiconductor memory (memory card, IC card) or the like can be considered. Further, the control program may be acquired via a so-called Internet or the like.

It is a block diagram which shows the structure of an image pick-up element. It is a block diagram which shows the structure of an image process part. It is a block diagram which shows the structure of a low-pass filter. It is a flowchart which shows operation | movement of a low-pass filter. It is a flowchart which shows operation | movement of the level 1 filter of a X direction. It is a block diagram which shows the structure of a high pass filter. It is a block diagram which shows the 1st modification of an image process part. It is a block diagram which shows the 2nd modification of an image process part. It is a block diagram which shows the structure of an aberration correction part. It is a flowchart which shows operation | movement of an aberration correction part. It is a flowchart which shows the calculation process of the scale value in an aberration correction part. It is a block diagram which shows the structure of the conventional image processing apparatus.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Imaging device, 2 Imaging part, 21 Imaging element, 22 Lens, 23 Motor, 24 Driver, 3 Image processing part, 32 Aberration correction part, 31 Gain adjustment part, 33 Low-pass filter, 34 High-pass filter, 35 Image composition part

Claims (11)

An image acquisition means for acquiring a visible light image and an invisible light image corresponding to the visible light image;
An image processing apparatus comprising: noise reduction means for reducing noise in the visible light image using the invisible light image. The noise reduction means is
First filtering means for applying a low-pass filter to the visible light image;
A second filtering means for applying a high-pass filter for the invisible light image;
The image processing apparatus according to claim 1, further comprising: a combining unit that combines the outputs of the low-pass filter and the high-pass filter.   The image processing apparatus according to claim 1, wherein the invisible light image is an infrared light image.   The image processing apparatus according to claim 1, wherein the invisible light image is an ultraviolet light image.   The image processing apparatus according to claim 1, wherein the low-pass filter is an edge-preserving low-pass filter.   The image processing apparatus according to claim 5, wherein the edge preserving low-pass filter removes noise from the visible light image while preserving edges detected from the invisible light image.   The image processing apparatus according to claim 1, further comprising an aberration correction unit that corrects aberrations of the visible light image and the invisible light image. A visible light image capturing means for capturing a visible light image based on a first spectral characteristic mainly sensitive to visible light;
An invisible light image capturing means for capturing an invisible light image based on a second spectral characteristic mainly sensitive to invisible light;
Aberration correction means for correcting aberration between the visible light image and the invisible light image;
An image pickup apparatus comprising: noise reduction means for reducing noise in a visible light image using the invisible light image. An image acquisition step of acquiring a visible light image and an invisible light image corresponding to the visible light image and captured with the same number of pixels as the visible light image;
A noise reduction step of reducing noise in the visible light image using the invisible light image. In a program for causing a computer to execute predetermined processing,
An image acquisition step of acquiring a visible light image and an invisible light image corresponding to the visible light image and captured with the same number of pixels as the visible light image;
A noise reduction step of reducing noise in the visible light image using the invisible light image. In a recording medium on which a program for causing a computer to execute predetermined processing is recorded,
An image acquisition step of acquiring a visible light image and an invisible light image corresponding to the visible light image and captured with the same number of pixels as the visible light image;
A recording medium on which a program having a noise reduction step of reducing noise of the visible light image using the invisible light image is recorded.

Images